A Sandra - univ-lorraine.frdocnum.univ-lorraine.fr/public/UPV-M/Theses/1996/Bintein... · 2016. 9....

AVERTISSEMENT

Ce document est le fruit d'un long travail approuvé par le jury de soutenance et mis à disposition de l'ensemble de la communauté universitaire élargie. Il est soumis à la propriété intellectuelle de l'auteur. Ceci implique une obligation de citation et de référencement lors de l’utilisation de ce document. D'autre part, toute contrefaçon, plagiat, reproduction illicite encourt une poursuite pénale. Contact : [email protected]

LIENS Code de la Propriété Intellectuelle. articles L 122. 4 Code de la Propriété Intellectuelle. articles L 335.2- L 335.10 http://www.cfcopies.com/V2/leg/leg_droi.php http://www.culture.gouv.fr/culture/infos-pratiques/droits/protection.htm

L ,aoqms s [n+ e6l]o

THESE

présentée à I 'université de Metz

pour I 'obtention du Diplôme de Doctorat

Spécial i té Toxicologie de I 'Environnement

Sylvain BINTEIN

OPTIMISATION ET VALIDATION DE

CHEMFRANCE: UN MODELE REGIONAL

DE FUGACITE NIVEAU III APPLIQUE A LA

FRANCE

Soutenue le 31. Octobre 1996 devant la Commission d'Examen:

MM.

Mlle

J. DEVILLBRS

J.M. JOUANYW. KARCHER

P. VASSEUR

Rapporteur

Rapporteur

ENEUOTTTTOUE UNNËRSITAIRE- -METZ

3310 (ûs I

sl b 3LIz,0I JAil. $97 -1996-

OI>vP

A Sandra

A mes parents

A mes frères et soeurs

Aux mineurs

2

Je remercie Monsieur J. Devillers de m'avoir accueilli dans son équipe et de

m'avoir fourni un sujet de recherche aussi porteur et intéressant que celui-ci.

Qu'il trouve ici I'expression de toute ma reconnaissance pour la compétence

scientifique et le soutien quotidien qu'il a su m'apporter.

Je remercie Monsieur J.M. Jouany et Monsieur W. Karcher pour m'avoir fait

I'honneur de juger cette thèse et de participer au jury. Je voudrais également

exprimer ma reconnaissance à Monsieur W. Karcher pour avoir permis le

financement par la Communauté Européenne de la plupart des études

présentées dans cette thèse.

Je remercie Mademoiselle P. Vasseur de m'avoir permis d'intégrer cette

formation doctorale et pour I'honneur qu'elle m'a fait de bien vouloir juger

mon travail.

Enfin, je voudrais exprimer ma reconnaissance à toutes les personnes qui

m'ont aidé à divers titres. Mes remerciement vont plus particulièrement:

- à mes collègues de laboratoire pour leur disponibilité et notanrment à D.

Domine pour son accueil chaleureux, son aide et ses produits régionaux,

- à tous les exilés du Nord et du Grand Nord pour leur amitié,

- aux "Bachelardiens" pour tous les moments de bonheur qu'ils m'ont

procurés.

1. INTRODUCTION

2. ESTIMATION

L'ADSORPTION

3. CHEMFRANCE

6. CONCLUSION

7. REFERENCES

SOMMAIRE

DE LA BIOCONCENTRATION ET DE

p.7

p.6

p. L0

p. 11

p. 15

p. 15

4. ' 'VALIDATION'' DE CHEMFRANCE

5. ETUDE COMPARATIVE DES EQUATIONS ESTIMANT LA

BIOCONCENTRATION P. 13

4

Cette thèse est constituée des articles suivants:

ARTICLE I: Bintein, S., Devillers, J. et Karcher, W. (1993). Nonlinear

dependence of fish bioconcentration on n-octanoVwater partition coefficient.

SAR QSAR Environ. Res.l, 29-39. p. L9

ARTICLE II: Bintein, S. et Devillers, J. (1994). QSAR for organic chemical

sorption in soils and sediments. Chemosphere2Sr llTl-1188. p. 31

ARTICLE III: Devillers, J., Bintein, S. et Karcher, W. (1995).

CHEMFRANCE: A regional level III fugacity model applied to France.

Chemosphere 30, 457 -476. P. 50

ARTICLE IV: Bintein, S. et Devillers, J. (1996). Evaluating the environmental

fate of lindane in France. Chemosphere 32r 2427-2440. p. 7l

ARTICLE V: Binrein, S. et Devillers, J. (1996). Evaluating the environmental

fate of atrazine in France . Chemosphere 32, 2441-2456. p. 86

ARTICLE VI: Devillers, J., Bintein, S. et Domine, D. (1996). Modeling the

environmental fate of atrazine. Présenté au 2llé^" Congrès de I'American

Chemical Society (Nouvelles-Orléans, Louisian e, 24-28 Mars, 1996).

Soumis dans Sulfur Triazine Herbicides: Risk Assessment (L. Ballantine, J.

McFarland et D. Hackett, Eds.). ACS Symposium Series, ACS Books,

Washington, D.C. p. 103

ARTICLE VII: Devillers, J., Bintein, S. et Domine, D.(1996). Comparison of

BCF models based on log P. Chemosphere (sous presse). p. ll7

I. INTRODUCTION

l-e, comportement d'une substance chimique dans I'environnement est

complexe. Il met en jeu de nombreux phénomènes liés à ses propriétés

physico-chimiques et aux caractéristiques des écosystèmes. Pour les produits

que I'on souhaite introduire sur le marché ou ceux déjà présents, I'estimation

de leur devenir dans I'environnement est primordiale. Le nombre

considérable de composés chimiques employés dans nos activités rendant

inconcevable la réalisation exclusive d'études in situ,, on fait de plus en plus

appel aux techniques de modélisation. La souplesse de ces approches permet

de concevoir des modèles globaux (e.g., Mackay et al., 1992) ou restreints à

un phénomène particulier (e.g., PRZM (Carsel et al., 1985)). Ces outils

mathématiques sont utilisés pour comparer le comportement de plusieurs

molécules, rechercher des informations avant la réalisation d'essais en

laboratoire ou sur le ferrain, ou encore identifier des compartiments ou des

secteurs à risques. Cependant, I'emploi de modèles dans les analyses de risques

exige une grande sûreté dans les résultats obtenus et une connaissance parfaite

de leurs limites d'utilisation. Dans ces conditions, I'objet de notre travail a été

d'optimiser et de valider CHEMFRANCE, un modèle régional de fugacité

niveau III appliqué à la France (Chancrogne, 1991). Nous devions élaborer de

nouvelles relations structure-bioconcentration et structure-adsorption

possédant un large domaine d'application, mais également confirmer les

fondements théoriques et empiriques de CHEMFRANCE en mettant en

parallèle les observations faites en laboratoire ou sur le terrain et les résultats

des simulations effectuées sur ordinateur.

2. ESTIMATION DE LA BIOCONCENTRATION ET DE

L'ADSORPTION (Articles I et II)

[æs modèles de distribution sont pour la plupart fondés sur la notion de

répartition compartimentale nécessitant I'utilisation de coefficients de partage.

Chacun d'eux, défini pour deux phases, gouveme la distribution à l'équilibre

du composé et donc de sa concentration finale dans les différents

compartiments de I'environnement. Ainsi, pour estimer le potentiel

d'accumulation des molécules dans les sols, les sédiments et les matières en

suspension, on utilise le coefficient d'adsorption (i.e., Kp). De même, le

facteur de bioconcentration (i.e., BCF) permet d'évaluer I'accumulation des

substances chimiques chez les êtres vivants par des voies non-alimentaires. La

plupart des études sur la bioconcentration sont réalisées sur le poisson, compte

tenu de son importance économique et de la disponibilité de tests standardisés.

De ce fait, cet organisme est fréquemment utilisé pour représenter le

compartiment biotique dans les modèles de distribution. Le Kp et le BCF

peuvent être déterminés expérimentalement ou par des relations de type

structure-activité (i.e., QSAR). Depuis le début des années 80, I'importance

considérable de ces deux paramètres a engendré la publication d'une multitude

d'équations (Lyman et al., 1990; ECETOC, 1995; GÛsten et Sabljic' 1995).

En dépit de ce fait, des problèmes de modélisation restaient en suspens. La

première partie de notre travail a donc consisté à établir de nouvelles

équations possédant un domaine d'application plus étendu. Pour estimer la

bioconcentration, les modèles régressifs linéaires entre les transformations

logarithmiques des BCF et des coefficients de partage n-octanolleau (i.e., log

P) sont couramment utilisés. Cependant, pour les substances très hydrophobes

(log P > 6), ils ne sont plus applicables (Banerjee et Baughman, 1991). Cette

cassure dans la relation linéaire entre une activité biologique et le caractère

hydrophobe des molécules a été soulignée dans diverses études (Devillers et

Lipnick, 1990; Hansch et Leo, 1995). En pharmacologie.et en toxicologie, ce

problème de linéaritê a étê, résolu grâce à I'utilisation de modèles paraboliques

et bilinéaires (Kubinyi, 1993; Hansch et Leo, 1995). L'article I présente

l'aptirude de ces techniques non-linéaires à modéliser la bioconcentration chez

le poisson. Dans un premier temps, une analyse bibliographique a été réalisée

pour obtenir de nombreuses valeurs de BCF et de log P. Une banque de

données contenant 154 molécules dont les valeurs de BCF suivaient des

critères de sélection précis a étê constituée. A partir de cet échantillon, trois

équations (i.e., linéaire, parabolique, bilinéaire) reliant le BCF au log P ont

été développées. L'analyse des paramètres statistiques et des résidus

correspondant à ces relations a mis en évidence, le faible pouvoir prédictif du

modèle linéaire dans notre étendue de log P (i.e., l,l2à 8,60) etla supériorité

du modèle bilinéaire face à l'équation parabolique. Un échantillon test

constitué de 29 molécules a permis de confirrner ces conclusions et d'étendre

la validiré de la relation bilinéaire aux substances possédant une valeur de log

P comprise entre 0,39 et 9,50.

Les problèmes de modélisation du coefficient d'adsorption sont différents. En

effet, la trop grande spécificité et I'absence de paramètres prenant en compte

le degré d'ionisation des acides et des bases correspondent aux principaux

défauts des équations existantes. La finalité de l'étude décrite dans I'article

II était donc de pallier ces problèmes en élaborant un modèle général

intégrant le caractère hydrophobe et le potentiel d'ionisation des substances

(i.e., log P et pKa, respectivement) ainsi que le pH et le pourcentage de

carbone organique (i.e., VoOC) pour décrire le substrat. Ce modèle a étê,

construit à partir de 229 valeurs de Kp corespondant à 53 composés et a été,

testé sur 500 Kp mesurés pour 87 molécules. Ces données ont été extraites

d'articles originaux et non de compilations. Trois étapes ont été nécessaires

pour obtenir une équation applicable aux composés ionisés et non-ionisés.

Dans un premier temps, une relation intégrant le log P et le VoOC a confirmé

la nécessité d'introduire des facteurs correctifs intégrant le degré de

dissociation des molécules. En effet, il est admis que pour les substances

capables de former des ions (i.e., acides, bases) des phénomènes de répulsion

ou d'affinité avec les colloides du substrat chargés négativement induisent,

respectivement, une baisse du coefficient d'adsorption pour les substances

présentant une charge négative et une augmentation pour celles chargées

positivement (Bailey et a1.,1968; Jafvert, 1990). Pour prendre en compte ces

phénomènes, le pKa des molécules et le pH de la solution des sols ou des

sédiments ont été incorporés dans une seconde équation. Leur intégration a

provoqué un net accroissement de la qualité des résidus pour les composés

ionisés, tout en conservant un haut pouvoir prédictif pour les autres

substances. Cependant, une analyse précise des résidus a mis en évidence une

sous-estimation générale des valeurs de Kp pour les bases. Cette tendance

s'explique par le fait que I'adsorption de ces molécules dépend de I'acidité de

surface inférieure de deux unités de pH à celle de la solution. L'intégration de

cette observation empirique dans notre équation a permis d'obtenir une

distribution aléatoire des résidus pour les bases. La supériorité de cette

dernière équation a été confirmée par les simulations réalisées sur notre

échantillon test, qui ont également permis d'étendre son domaine d'application

à de nouvelles classes de substances chimiques et à des conditions pédologiques

très variées.

3. CHEMFRANCE (Art ic le I I r)

Les modèles de distribution peuvent être ponctuels ou globaux. Les premiers

perrnettent d'obtenir, en fonction de leur degré de complexité, des résultats

précis et une bonne compréhension des phénomènes étudiés. Cependant, la

masse d'information nécessaire à leur emploi limite leur domaine

d'application. A I'inverse, la finalité des modèles globaux est d'estimer le

comportement des molécules dans un environnement représenté par plusieurs

compartiments. Pour cela, un compromis entre Ia réduction des informations

nécessaires à leur fonctionnement et la pertinence des résultats fournis doit

être trouvé. Ces outils de simulation sont fondés, en quasi-totalité, sur le

concept de fugacité (Mackay, 1991; Cowan et al., 1995). Parmi eux, on

distingue les modèles régionaux et les modèles d'évaluation employant un

"monde unité" représentant un environnement hypothétique. [æs premiers

tentent de s'approcher le plus possible d'une situation réelle afin d'obtenir des

informations comparables aux données expérimentales. Les seconds

perïnettent uniquement de comparer les molécules entre elles ou d'identifier

les principaux processus mis en jeu. L'article III présente CHEMFRANCE,

un modèle régional de fugacité niveau III appliqué à la France. Le niveau III

l0

signifie que le système est à l'état stable (i.e., flux d'entrée = flux de sortie)

mais que des transferts entre les compartiments sont possibles (e.9.,

volatilisation, ruissellement, lessivage). CHEMFRANCE comprend six

compartiments globaux (i.e., air, eaux de surface, sol, sédiments, eaux

côtières, eaux souterraines). Les quatre premiers sont constitués de sous-

compartiments (e.g., poissons, matières en suspension). CHEMFRANCE

perrnet de simuler, pour les quatre saisons, le comportement des molécules en

France et dans 12 régions françaises définies en fonction de leurs

caractéristiques pédologiques, climatiques et hydrologiques. Les paramètres

environnementaux (e.g., pH du sol, volume des eaux de surface)

indispensables au fonctionnement de CHEMFRANCE ont été déterminés pour

chacune de ces régions. Ainsi, I'utilisateur doit fournir simplement les

propriétés physico-chimiques et les quantités d'émissions de la molécule à

analyser pour obtenir la concentration, la quantité absolue, les cinétiques

d'élimination et d'échange pour chaque compartiment...

4. "VALIDATION" DE CHEMFRANCE (Articles M VI)

Afin d'obtenir un outil de simulation efficace et fiable, tout logiciel doit être

vérifié et testé (Devillers et al., 1992, 1995; Fredenslund et al., 1995; Vincent

et al.,1996). Pour les modèles de distribution, la validation des informations

produites présente le niveau de difficulté le plus élevé. La complexité de cette

procédure est due à la rareté mais surtout à la variabilité spatiale et

temporelle des données expérimentales utilisées comme éléments de

1t

comparaison. Dans ces conditions, la validité d'un modèle ne pouna jamais

être pleinement affirmée. On pourra simplement confirmer la pertinence des

résultats obtenus en démontrant les concordances entre observations et

prédictions. Dans ce contexte, le pouvoir de simulation de CHEMFRANCE a

été évalué sur le lindane (Article IV) et I'atrazine (Articles V et VI). Du

fait de l'utilisation massive de ces pesticides, de multiples études ont étê

entreprises afin de connaître avec précision tous les aspects de leur

comportement dans I'environnement. Une synthèse de ces informations nous a

permis d'établir un profil environnemental assez précis de ces molécules. Une

fois cette base de travail établie, les données indispensables au fonctionnement

du programme (i.e., propriétés physico-chimiques, tonnages émis) ont été

déterminées le plus précisément possible pour faciliter la comparaison entre

les valeurs expérimentales et celles calculées par CHEMFRANCE. Quatre

scénarios intégrant les méthodes d'application de ces substances actives et

leurs déversements dans les eaux de surface ont été simulés. Cette approche

possède deux intérêts. D'une part, elle permet d'identifier les risques de

contamination et les principaux processus mis en jeu en fonction des

compartiments d'émission et d'autre part, d'obtenir une variabilité dans les

résultats. De fortes similitudes entre les résultats extraits de ces simulations et

ceux obtenus en laboratoire ou sur le terrain ont été observées. Ainsi, le

lindane est présent principalement dans les sols, les sédiments et les éléments

biotiques du fait de ses capacités d'adsoqption et de bioconcentration. A

I'inverse, I'atrazine se retrouve principalement dans les milieux aquatiques

bien que la contamination atmosphérique ne puisse pas être êcartée. Cette

molécule est faiblement adsorbée par les sols induisant par conséquent un fort

t2

potentiel de ruissellement et de lessivage. Enfin, elle est très peu

bioaccumulée. Une analyse sensitive portant sur le temps de demi-vie de

l'atrazine dans les sols est présentée dans I'article VI. Cette étude montre

que la concentration dans les sols et les eaux souterraines et les processus de

volatilisation au niveau du sol, de ruissellement et de lessivage sont les

paramètres les plus sensibles aux variations de la vitesse de dégradation de

l'atrazine dans les sols.

5. ETUDE COMPARATIVE DES EQUATIONS ESTIMANT LA

BIOCONCENTRATION (Art ic le VII)

Les modèles de type QSAR peuvent s'avérer des outils précieux pour évaluer

les risques écotoxicologiques s'ils sont performants et surtout utilisés à bon

escient. n est donc indispensable de connaître précisément leurs limites

d'utilisation. Dans ce contexte, nous avons entrepris de comparer différents

modèles perrnettant d'estimer la bioconcentration des molécules organiques.

Tous les modèles testés utilisaient le log P comme descripteur moléculaire.

La sélection des équations a étê motivée par une étude d'occurrences effectuée

à partir de rapports élaborés pour des organisations officielles (e.g., OCDE,

EPA, UE). L'intégration de ces équations dans les modèles de distribution a

également influencé notre choix. La réalisation de nos études comparatives

soulevait un certain nombre de problèmes méthodologiques. En effet, les

données expérimentales sur la bioconcentration des molécules sont

relativement peu nombreuses et ont donc largement été utilisées pour élaborer

13

les équations à comparer. Or, lorsque I'on veut estimer le pouvoir prédictif

d'un modèle de type QSAR, il est indispensable d'utiliser un échantillon test

constitué de molécules n'appartenant pas à l'échantillon d'apprentissage. La

solution optimale aurait été de mesurer des valeurs de BCF pour de nouvelles

substances. Dans le cadre de notre travail, pour des raisons de temps et de

coût, une telle alternative ne pouvait être envisagée. Nous n'avons donc pas

cherché à sélectionner une seule valeur par composé mais plutôt, dans la

mesure du possible, à retenir un ensemble de résultats expérimentaux

cohérents permettant de fournir un profil de bioconcentration pour une

molécule donnée. Une étude bibliographique visant à récolter un grand

nombre de valeurs expérimentales a êté réalisée à partir de publications

originales et non de compilations afin d'obtenir des informations sur les

conditions expérimentales. L'article VII présente une étude comparative

réalisée sur sept équations linéaires ou non-linéaires peffnettant d'estimer le

BCF à partir du log P (Veith et al., 1979, 1980; Mackay, 1982; Connell et

Hawker, 1988; Isnard et Lambert, 1988; Nendza, l99l; Bintein et al., 1993).

Cette analyse a été effectuée à partir d'une banque de données constituée de

342 valeurs de BCF provenant d'essais dynamiques et 94 déterminées par

essais statiques ou semi-statiques. Pour des valeurs de log P inférieures à 6,

les différents modèles testés donnent des résultats équivalents. Par contre,

pour les substances très hydrophobes (i.e., log P > 6), les relations linéaires

ne sont plus applicables et sont nettement devancées par les équations non-

linéaires. Parmi celles-ci, notre équation possède le meilleur pouvoir prédictif

et le plus large domaine d'application.

t4

6. CONCLUSION

En élaborant de nouvelles relations structure-bioconcentration et structure-

adsorption nous avons accru le domaine d'application et le pouvoir de

simulation de CHEMFRANCE. Les études réalisées sur le lindane et I'atrazine

ont démontré le haut potentiel de simulation de CHEMFRANCE et ont permis

d'effectuer une synthèse du comportement dans I'environnement de ces

pesticides. Cependant, nous envisageons de poursuivre ce processus de

validation avec d'autres molécules aux propriétés physico-chimiques variées

pour définir d'une façon plus précise les limites d'utilisation de ce logiciel.

Nous prévoyons également d'intégrer dans CHEMFRANCE la méthode SIRIS

(Vaillant et al., 1995) ainsi que des informations écotoxicologiques afin de

mieux appréhender les risques (éco)toxicologiques et de créer un outil d'aide

à la décision plus facile à utiliser.

7. REFERENCES

Bailey, G.W., White, J.L. et Rothberg, T. (1968). Adsorption of organic

herbicides by montmorillonite: Role of pH and chemical character of

adsorbate. SoiI Sci. Soc. Amer. Proc.32,222-234.

Banerjee, S. et Baughman, G.L. (1991). Bioconcentration factors and lipid

solubility. Environ Sci. Technol. 25, 536-539.

Bintein, S., Devillers, J. et Karchef, W. (1993). Nonlinear dependence of fish

bioconcentration on n-octanol/water partition coefficient. SAR QSAR

Environ. Res. 1, 29-39.

15

Carsel, R.F., Mulkey, L.A., Lorber, M.N. et Baskin, L.B. (1985). The

pesticide root zone model (PRZM): A procedure for evaluating pesticide

leaching threats to groundwater. Ecol. ModellinS 30' 49-69.

Chancrogne, A. (199L). CHEMFRANCE: Un modèle de fugacité niveau III

appliqué à la France. D.E.A. Toxicologie de l'Environnement. Centre des

Sciences de I' Environnement, Metz, p. 43.

Connell, D.W. et Hawker, D.W. (1983). Use of polynomial expressions to

describe the bioconcentration of hydrophobic chemicals by fish. Ecotoxicol.

Environ. Safery 16, 242-257.

Cowan, C.E., Mackay, D., Feijtel, T.C.J., van de Meent, D., Di Guardo, A.,

Davies, J. et Mackay, N. (1995). The Multi-Media Fate Model: A Vital Tool

for Predicting the Fate of Chemicals. SETAC Press, Pensacola, p. 78.

Devillers, J. et Lipnick, R.L. (1990). Practical applications of regression

analysis in environmental QSAR studies. Dans, Practical Applications of

Quantitative Structure-Activity Relationships (QSAR) in Environmental

Chemistry and Toxicology (W. Karcher et J. Devillers, Eds.). Kluwer

Academic Publishers, Dordrecht, pp. 129-143.

Devillers, J., Bintein, S. et Domine, D. (1995). Evaluation of USES l -0.

French Contribution to the Development of EU Risk Assessment Guidnnces

for New and Existing Substances. Ministère de I'Environnement

(BSPC)/CTIS, p. 39.

Devillers, J., Domine, D. et Binteiil, S. (1992). Study of Four Exposure

Models applied to Ten Chemicals Substances. Ministère de I'Environnement

(BSPC)/CTIS, p. 51.

ECETOC (1995). The Role of Bioaccumulation in Environmental Risk

Assessment: The Aquatic Environment and Related Food Webs. Technical

Report No. 67, p. ll2.

Fredenslund, F.C., Severinsetr, M. et Andersen, M.B. (1995). Evaluntion of

the SimpleBox Mod.el for Danish Conditions. Environmental Project n" 307.

Ministry of Environment and Energy, Danish Environmental Protection

Agency, Copenhagen, p. 96.

Gûsten, H. et Sabljic, A. (1995). QSARs for soil sorption. Dans, Overview of

16

Structure-Activity Relationships for Environmental Endpoints. Part I:

General Outline and Procedure. Report of the EU-DG-XII Project QSAR for

Predicting Fate and Effects of Chemicals in the Environment. Contract #

EV5V-CT92-0211 . ChaPter III.

Hansch, C. et Leo, A. (1995). Exploring QSAR. Fundamentals and

Applications in Chemistry and Biology. ACS, Washington, p. 557.

Isnard, P. et Lambert, S. (1988). Estimating bioconcentration factors from

octanol-water partition coefficient and aqueous solubility. Chemosphere 17,

2l -34.

Jafvert, C.T. (1990). Sorption of organic acid compounds to sediments: Initial

model development. Environ. Toxicol. Chem.9, 1259-1268.

Kubinyi, H. (1993). QSAR: Hansch Analysis and Related Approaches. VCH,

Weinheim, p. 240.

Lyman, W.J., Reehl, W.F. et Rosenblatt, D.H. (1990). Handbook of Chemical

Properry Esfimation Methods. ACS, Washington.

Mackay, D. (1982). Correlation of bioconcentration factors. Environ. Sci.

1 frchnol. 16, 274-278.

Mackay, D. (1991). Multimedia Environmental Models. The Fugaciry

Approaclr. Lewis Publishers, Chelsea, p. 257.

Mackay, D., Paterson, S. et Shiu, W.Y. (1992). Generic models for evaluating

the regional fate of chemicals. Chemosphere 24, 695-717.

Nendza, M. (1991). QSARs of bioconcentration: Validity assessment of log

Pow[og BCF correlations. Dans, Bioaccumulation in Aquatic Systems (R.

Nagel et R. Loskill, Eds.). VCH, Weinheim,pp.43-66.

Vaillant, M., Jouany, J.M. et Devillers, J. (1995). A multicriteria estimation

of the environmental risk of chemicals with SIRIS method. Toxicol. Model

l , 57 -72.

Veith, G.D., DeFoe, D.L. et Bergstedt, B.V. (1979). Measuring and

estimating the bioconcentration factor of chemicals in fish. ,I. Fish. Res. Board

Can.36' 1040-1048.

Veith, G.D., Macek, K.J., Petrocelli, S.R. et Carroll, J. (1980). An evaluation

of using partition coefficients and water solubility to estimate

T7

bioconcentration factors for organic chemicals in fish. Dans, Aquatic

Toxicology, ASTM STP 707 (J.G. Eaton, P.R. Parrish et A.C. Hendricks,

Eds.). American Society for Testing and Materials, pp. 1L6-129.

Vincent, R., Devillers, J., Bintein, S., Sandino, J.P. et Karcher, W. (1996).

Evaluation of EASE 2.0. Assessment of Occupattonal Exposure to Chemicals.

INRS/CTISÆCB,p. 48.

18

ARTICLE I

t9

SAR ad QSAR in Enuironmmtal Research, Vol. t. pp. 2F39

Reprints available direclly from the publisher

PhotocopyinS pcrmitted by license only

@ 1993 Gordon and Breach Science Publishen S.A.

Printed in thc United States of Amedca

NONLINEAR DEPENDENCE OF FISHBIOCONCENTRATION ON z-OCTANOL/WATER

PARTITION COEFFICIENT

S. BINTEIN and J. DEVILLERST

CTIS.2l rue de la Bannière,69003 Lyon (France)'

W. KARCHER

commission of the European communities, Joint Research centre,ISPRA Establishment, I-21020 Ispra Varese (Italy)'

(acc'ePted JulY 28, 1992)

The log1og relationship between the bioconcentration tendency of organic chemicals in fish and the

2 -octaiol/îater partiti;n coefficients breaks down for very hydrophobic compounds. The use of para-

bolic and bilinear models allows this problem to be overcome. The QSAR equltion log BCF:

0 . 9 1 0 l o g P - l . g 7 5 l o g ( 6 . 8 l 0 - 7 P + l ) 0 . 7 8 6 ( n : 1 5 4 : r : 0 . 9 5 0 ; s : 0 . 3 4 7 ; F : 4 6 3 . 5 l l w a s f o u n d t obe a good predictor of bioconcentration in fish.

KEY WORDS: Bioconcentration; BCF; fish; bilinear model; linear model; parabolic model.

INTRODUCTION

Bioconcentration is the process of accumulation of chemicals by organisms through

nondietary routes.r In aquatic ecosystems, the bioconcentration factor (BCF) of an

organic chemical is defined as the ratio of its concentration in a target organism to

that in water at steady state.2 Typically, fish are the target organisms of BCFassessments due to their importance as a human food source and the availability ofstandardized testing protocols.l The most common method for estimating chemicals'BCFs consists of establishing correlations between BCFs and hydrophobicity of thechemicals. The majority of these are obtained from linear regression models betweenthe log transformations of the BCF values and the n-octanol/water partitioncoemcients (log P) of the chemicals.3-6 The regression equations have the followinggeneral form:

logBCF-a logP+b

* Author to whom all correspondence should be addressed'

29

(1)

20

S. BINTEIN, J. DEVILLERS AND rril. KARCHER

where a and b are constants. In most published BCF equations, the slope (a) is < 1and the intercept (b) is negative.6 Equation I breaks down for strongly hydrophobicchemicals (log P > 6).' The break in the linear relationship between a biologicalactivity (BA) and the hydrophobic character of the organic molecules has beenstressed by different authors in numerous QSAR studies.s Hansch initiated the useof a simple parabolic models in log P (Eq. 2) to overcome this problem.

log BA -- a log P + b (log P)2 + c

Kubinyie-r2 subsequently proposed a better model (Eq. 3) that adequately accountedfor the fact that the biological activity and partition coefficient initially vary in alinear fashion on a log-log scale, reaching an optirnum value, followed by a seconddecreasing linear portion.

log BA : alog P - b log(pP * l) + c

This bilinear model (Eq. 3) has been successfully used in many drug design andenvironmental QSAR studies.r3-r8

The aim of this paper is to compare the performance of the parabolic and bilinearmodels to overcome the "cut-off" problem encountered in the estimation of BCFvalues from the n-octanol/water partition coefficient (log P).

MATERIALS AND METHODS

Data selection

BCF values for 154 chemicals (Figure 1) were obtained from an extensive compila-tion.le Data were selected according to the following criteria:

l) Only experimental data, measured in whole fish (wet weight), and obtainedunder flow-through conditions, were used.

Substiurtcd benzencsPosticidesPCB(12)+ PBB( l )AliphaticsPOD(7)+PCDF(2)PACMisoellaneous

(2)

(3)

Figme I Data set constitution. (PCB = polychlorobiphenyls; PBB: potybromobiphenyls; PCDD =polychlorodibenzo-p-dioxins; PCDF = polychlorodibenzofurans; PAC: polyaromatic compounds).

2l

NONLINEAR MODEL FOR BCF 31

2,) Datawere included only if a steady state was reached or when BCF values lvere

obtained from the kinetic method.

3) BCF data were rejected if contamination by food and/or presence of adsorbents

(e.g., suspended sediments) was suspected.

4) Only studies using freshwater fish were considered. Five families \4'ere re-

presented in the selected data:

- Salmonidae: rainbow trout, whitefish,

-Cyprinidae: fathead minnow, goldfish, carps (from different geographical

origins), zebrafish, bleak, and topmouth gudgeon

- Centrarchidae: bluegill sunfish,

- Poeciliidae: guPPY'- Cyprinodontidae: American flagfish and killifish'

Fish species distribution in the study is summarized in Figure 2' The log- P values

were ietrieved from an extensive bibliographical review aimed at preferentially

selecting data obtained by direct methods (e.g', shake-flask, generator column)'

Moilel deuelopment

Description and calculation procedures for the bilinear model can be found in

numerous comprehensive papôrs.e-12 Briefly, the bilinear model is derived from the

McFarland multicompartmental model *tti.tt allows the estimation of the prob-

uùif,y p of a molecuÈ crossing aqueous-lipid interfaces and reaching its receptor

site. Thus, if we consider a hypothetical biological system made up of alternating

aqueous phases A and lipid (membrane) phases L (Figure 3), the probability of a

-29

Figure 2 Number of BCF values per species'

l8

IIEINtrEE

Rainbow troutFathead minnowZbrafishTopmouth gudgeonBluegill sunfishGuppyOther species Cl)

22

32 S. BINTEIN, J. DEVILLERS AND W. KARCHER

Figure 3 The McFarland hypothetical biological systemdistribution process).

il(k ' and k, are the rate constants of the

molecule leaving Ao and entering into L, is:

P o , t : - O ' -

kr+k,

In the same way, the probability that a molecule goes from L, to A, is

k2Pr,z: i r i t ,

(s)

If both parts of the fractions in Eqs. 4 and 5 are divided by k, and if krlk2 is replacedby P (the partition coefficient), we obtain:

(4)

Therefore, the probabilityfrom .40 to An is given by

PPo, t : p + I

IP r .z : F+ f

that a molecule goes from Ao to A, andEqs. 8 and 9 respectively.

(6)

(7)

more generally

PPo,z: Po. r X Pt,z: --------------^(P + l) '

PnlzP o . n : P o . r x P t , z x " ' x P n - t . n : 1 p _ r l ; n ( 9 )

If we consider first that the biological activity (BA) of a molecule is a function of itsintrinsic activity and of its probability of reaching the receptor site, and second thatthe intrinsic activities of homologous series of chemicals are identical, we can write:

BA:cons tan txPo(P + l)2^

The logarithmic transformation of Eq. l0 leads to Eq. 11.

log BA : n lo9P -2n log(P + 1) + c (11)

Symmetrical curves with linear ascending and descending sides and a parabolic partwithin the range of optimum lipophilicity result from Eq. 11.

(8)

(10)

23

^o

kt

L I

k2

u

A2

L ' l

kt

R

t 2

NONLINEAR MODEL FOR BCF

Figure 4 Hypothetical biological system for the bilinear model (kr and k, are the rate constants of thedistribution process).

The bilinear model is derived from the McFarland model (Eq. 1l) by taking intoaccount the different volumes of the aqueous and organic phases in a biologicalsystem comprising four phases (Figure 4). The general equation of the bilinear model(Eq. 3) applied to BCF and log P is:

log BCF : alog P - blog(pP + l) + c

33

(r2)

where p is the ratio volume between the lipid and the aqueous phases. For small Pvalues, (PP + 1) is approximately equal to one, therefore log(BP * 1) equals zero.For large P values (llP + 1) is nearly equal to flP, therefore log(fP + l) is propor-tional to log P.

The parameters a, b and c are linear terms, which can be calculated by linearmultiple regression analysis, B is a nonlinear parameter, which is generally estimatedby a Taylor series iteration method.2o

Calculations were performed as described by Kubinyi and Kehrhahn,ro fromsoftware written in Turbo basic and running on an IBM PC compatible.

RESULTS AND DISCUSSION

Figure 5 depicts the ability of the linear (Eq. 13), parabolic (Eq. la) and bilinear(Eq. 15) models to describe the set of 154 BCF values from log P.

log BCF:0 .516 log P + 0 .576

n -- 154; r :0.772; s : 0.702; F :224.32

logBCF : -0.164 (log P)2 +2.0591og P -2.592

n :154 ; r :0 .914 ; s :0 .450 ; F :382 .09

log BCF : 0.9101og P - I.975log (6.8 10-7P + 1) - 0.786

n :154 : r :0 .950 ; s :0 .347 ; F :463 .51

From the above and from Figure 5, it is obvious that Eq. 15 is the best regressionmodel and that the linear solution given by Eq. 13 is not valid in our range of log P

values (i.e. 1.12-8.60). This is confirmed by inspection of the statistical parameters

of Eq. 13 and the graphical analysis of the residuals in Figures 6 and 7. Furthermore,lve can note the unsatisfactory behaviour of the residuals in Figure 6 which aredisplayed according to a "parabolic band". This confirms the possibility of using a

(13)

(14)

(ls)

24

34 S. BINTEIN, J. DEVILLERS AND IV. KARCHER

log BCF6

5

4

3

2

l { . :

0

. l

r 2 3 4 5 6 7 8 9 1 0

Figure 5 Relationships between log BCF and log P'

log P

Figure 6 Distribution of the residual values obtained with the linear model (Eq' 13)'

45

4035

3025?!

l 5

l0

50

0 0.t 02 03 0.4 05 0.6 0.? 0.8 0.9 I t.l l, 13 1.1 15 1.6 lJ l.t t'9 >2

Residuals

Figure 7 Residual analysis of the linear model (Eq' l3)'

.rEq.tl

E q . 1 4

E q . 1 5

log P

5

4

3

26r€

OE- l

-2

-3

.5

4

û . 3(JA 1n -c6 r

= to

ù 0

- l

-2

-3t 2 3 4 5 6 7 8 9

25

quadratic term in the BCF model in order to increase its statistical qualities andpredictive power.2o Thus, introduction of (rog p)z in the moder (Eq. ri) reads to abetter distribution 9f1he residuals (Figures

g ano 9). In the same way, the high qualityof the bilinear model (Eq. l5) is confirmed by analysis of the residuat fËigures toand 1l) which reveals that no chemicals .un

-b. considered as outliers. io Ëstimate

the predictive power of the bilinear model (Eq. l5) a testing set of 29 chemicals hasbeen constituted (Table I). In Table I, experiméntai BCF values were obtained underflow-through or static conditions and generally did not satisfy all the other constraintsdefined for the selec.tion of the training set. Comparison of ihe expected BCF valuesobtained from the different models (i.e., Eq. l3 t; l5) shows that the bilinear model(Eq. 15) is the most suitable to describe the-BCF behaviour of the chemicals listed inTable I.

Wh-ile it is recognised that the bioconcentration of organic chemicals in fish is acomplex phenomenon involving numerous biotic and àbiotic factors, our resultsunderline the ability of the bilinear model of Kubinyie-r2 to simulate this process in


r 2 3 4 5 6 7 8 9log p

Figure 8 Distribution of the residual values obtained with the parabolic model (Eq. l4).

Frequencies50

4540

35

30

25

20l5

l0

5

00 0.1 0, 03 0.4 0.5 0.6 0.7 0.8 o.Ij,oljor.r l3 t.4 t.5 t.6 t.7 t.8 t.e >2

Figure 9 Residual analysis of the parabolic model (Eq. la).

26


1 2 3 4 5 6 7 8 9log P

Figure l0 Distribution of the residual values obtained with the

Frequencies5045

û

353025

nl 5

l0

50

0 0.1 02 0.3 0.4 05 0.6 0J 0.8 0.9 I l.lResiduals

a rather simplistic way. If only a few data points are available or if the log P values

are not too high, the linear models3-6 are the most suitable to describe the bio-

concentration in fish due to their simplicity. However, if enough data points are

present and if the log P values vary over a wide range, the parabolic model and

preferentially the bilinear model are more relevant. Thus, it is interesting to note that

if we consider the range of log P values in our training and testing sets, the bilinear

model (Eq. 15) allows the prediction of the bioconcentration factor ôf organic

chemicals having log P values between 0.39 and 9.50.Chemicals whichlre released to the environment as the result of a variety of

human-related activities migrate through the ecosystems according to their physico-

chemical properties (e.g., vapour pressure, hydrosolubility). Numerous models vary-

ing in compiexity are-available to evaluate such behaviour.32 Among them, the

fuiacity models of Mackay33 provide a convenient and accurate simulation tool for

r*iorur. and risk analysis. In the classical fugacity models,33 the bioconcentration

t.8 1.9 >2

J

2(,

t a' o

oit- l

-3

bilinear model (Eq. l5).

1.6

Figure ll Residual analysis of the bilinear model (Eq' l5)'

n

NONLINEAR MODEL FOR BCF 37

Table I Comparison between experimental and calculated BCF values for 29 chemicals.

No. Chemical log P log BCF.,, (ref) Calculated log BCF

Eq. 13 Eq. 14 Eq. 15

I2J

456789

l0l lt2l 3l 4l 5l 6l 7t 819202 l2223242526272829

2-(t-Butoxy) ethanolt-Butyl isopropyl etherOctachloronaphthalenePigment, monoazo"Pigment, monoazo"Anilinet-Butylphenyldiphenyl phosphate2-Chloroaniline3-Chloroaniline4-Chloroaniline2-Chloronaphthalene2,4-Dichloroaniline3,4-Dichloroaniline1,4-Dichloronaphthalene1,8-Dichloronaphthalene2,3-Dichloronaphthalene2,7-Dichloronaphthalcnc2-Nitroanil ine3-Nitroaniline4-Nitroanilinee3,7,8-TCDD3,3',4,4' -T etr achlorobi phenyl3,3',4,4' -T etr achlorodiphen yl ether2,4,S-Trichlorobiphenyl2,4,5-Trichlorodiphenyl ether3,4,5-TrichloroveratroleTetrachloroveratrole1,3,7-TrichloronaphthaleneTriphenyl phosphate

0.39' -0.22 (2ll2.14^ 0.76 (2ll8.40h 2.52 (22)9.50 0.70 (23)9.30 0.s0 (23)0.90 0.41(24)5.12 3.04 (25)1.90 r.18 (24)1.88 t.06 (24)1.83 0.91(2414.19 3.63 (26)2.7e t.98 (2412.79 t.48 (24)4.88 3.36 (26)4.4t 3.79 (26)4.71 4.M (2614.8r 4.M (261r.78 0.9t (24)l.3l 0.92 (241r.3l 0.64 (24)6.42d 3.90 (271s.82 4.s9 (28)s.78 4.51 (28)s.sl 4.26 (28)s.44 4.18 (28)4.60 3.s0 (29)s.80 4.40 (29)5.59 4.43 (2613.90" 2.76 (25)

0.781.684.915.485.371.043.22r.561.551.522.742.022.023.092.853.013.061.491.251.253.893.583.563.423.382.953.573.462.59

- l .8 l1.063.1 32 . 1 72.37

-0.873.650.730.700.633.16r .88t .883.553.303.473.520.55

- 0 . 1 8-0 .18

3.873.843.833.773.763.413.833.792.94

-0.43l . l 62.44r.281.490.033.800.940.920.883.021.751.753.613.213.473.550.830.410.414.184.194.184.064.023.384.194.102.76

" mean of two values.b Ref. 26."Chemicals no.45 and 46 in Anliker et a1.,23 lTable | [part B] p.267).d Ref. 30." Ref. 31.

in the biota (i.e., fish) is estimated from a simple linear regression equation (Eq. l).This induces erroneous estimations for highly lipophilic chemicals. To overcome thisproblem, we propose to introduce in the fugacity models our bilinear model (Eq. l5)in order to stretch their domain of application.

References

1. Barron, M. G. (1990). Bioconcentralion. Enuiron. Sci. Technol. U' 1612-1618.2. Hamelink, J.L. (1977). Current bioconcentration test methods and theory. ln, Aquatic Toxicology and

Hazard Evaluation (F. L. Mayer and J. L. Hamelink, Eds.). ASTM STP 63f' pp. 149-161.3. Kanazawa, J. (1981). Measurement of the bioconcentration factors of pesticides by freshwater fish and

their correlation with physicochemical properties or acute toxicities. Pestic. Sci. 12' 417424.4. Schùûrmann, G. and Klein, W. (1988). Advances in bioconcentration prediction. Chëmosphere 17,

t55r-r574.

28


5. Isnard, p. and Lambert, s. (1988). Estimating bioconcentration factors from octanol-s'ater partition

coefficient and aqueous solubility. Chemosphere 17,21-34'6. Samiullah, y. (1990)- prediction o7 tlrc tnit onmenial Fate of Chemicals- Elsevier Science Publishers'

London, p.285.Z. Sun"rj"é,'S. and Baughman, G. L. (1991). Bioconcentration factors and lipid solubility. Enoiron'

Sci'

Technol. 25, 53ç539.8. Devillers, J. and Lipnick, R. L. (1990). Practical applications of regression analysis in environmental

eSAR studies. ln, p:rouià Àp)tirotiont of Quaniiiatioe Srru.cture'Actiuity Relatioys-hips (Q.SAR) in

Enoironmenral Chemfsçy ani'To-ricoloSy- (fr.Karcher and J. Devillers, Eds.). Kluwer Academic

Publishers, Dordrecht, pp' 129-143.9. Kubinyi, H. (1976). Quantitative structure-activity relationships'. IV' Non-linear dependence

of

biological activity on hydrophobic character: a new model. Arzneim. Forsch' Drug R9s' 74,-1991-1997 '

fO. XuUinyl, H ana i


29. Neilson, A. H., Allard, A. S.,R.eiland, S., Remberger, M., Târnholm, A., Viktor, T., and Landner, L.(1984). Tri- and tetra-chloroveratrole, metabolites produced by bacterial O-methylation of tri- andtetra-chloroguaiacol: An assessment of their bioconcentration potential and their effects on fishreproduction. Can. J. Fish. Aquat. Scr'. 41, 1502-1512.

30. Sijm, D. T. H. M., Wever, H, de Vries, P. J., and Opperhuizen, A. (1989). Octan-1-ol/water partitioncoefficients of polychlorinated dibenzo-p-dioxins and dibenzofurans: Experimental values determinedwith a stirring method. Chemosphere 19,263-266.

31. Bengtsson, B. E, Tarkpa, M., Sletten, T., Carlberg, G. E., Kringstad, A., and Renberg, L. (1986).Bioaccumulation and efrects of some technical triaryl phosphate products in fish and Nitoua spinipes.Enairon. Toxicol. Chem. 5, 853-861.

32. Cohen, Y. (1986). Organic pollutant transport. Enuiron. Sci Technol. 20,538-544.33. Mackay, D. (1991). Multimedia Enuironmental Models. The Fugacity Approach. Lewis Publishers,

Chelsea, p.257.

39

I

30

ARTICLE II

3l

@ Pergamon

0045-6535(94)E00s0-4

QSAR FOR ORGANIC CHEMICAL SORPTION IN SOILS AND SEDIMENTS

S. Bintein and J. Devillers*

CTIS, 2l rue de la Bannière, 69003 Lyon, France

(Rcccivcd in Gormrny 30 Novembcr 1993; rccoptcd lt hnurry lg4)

ABSTRACT

Sorption phenomena exert important influences on the environmental fate of anthropogenic

organic substances. Under these conditions, the aim of this study was to propose a general

QSAR model using the physicochemical properties of the molecules (i.e. log Ksy and pKa)

and some relevant properties of soils or sediments (i.e. pH and %OC) to estimate the sorption

behavior of both ionized and non-ionized chemicals. The proposed model was elaborated

from 229 Kp values rdcorded for 53 chernicals. The model was then tested on 500 other Kp

values obtained for 87 chemicals.

INTRODUCTION

fip fate of the organic chemicals introduced into the environment depends on a variety of

physicochemical and biological processes.l-2 Mattrematical models which attempt to integrate

these phenomena are widely used to predict the environmental transport and distribution of

the organic pollutants between the different compartments of the biosphere.S-7 Use of these

models requires a variety of abiotic and biotic parameters as inputs. Among them, the soil or

sediment sorption coefficient of chemicals (Kp) is one of the key input parameters in models

used to estimate the environmental mobility and fate of the pollutants.S Because its

experimental determination is time-consuming and expensive, estimated values based on

QSAR (Quantitative Structure-Activity Relationship) equations are now widely used.

fire distribution of organic chemicals between soils or sediments and water can be quantified

* Author to whom all corrcspondence should be addressed.tLTl

Chemosphete, Vol. 28, No. 6' p' ll7l-tltt' 1994Copyright {9 1994 Elscvicr Scicnoo Lrd

hinted in Great Britain. All righc rescwedm45-6535f94 $6.æ1{).00

32

l t72

through the use of the Freundlich isotherm:S=Kp*CN ( l )

where S is the sorbed concentration, C is the solution concentration, and Kp and N

(traditionally expressed as l/n) are empirical constants. The values of N commonly range

betrpeen 0.7 and 1.2.9The influence of the organic matter from soils and sediments on the sorption behavior of the

organic chemicals has been discussed in many studies.l0-l2 Consequently, the sorptionpartition coeffïcient Kp is generally related to ttre fraction of organic carbon (fsc) associated

with the sorbent to yield an organic-carbon-partition coeffTcient, Ksç,

Koc = (Kp'r 100)/VoOC (2)

Sometimes, sorption coefficients are nornalized to the organic matter content instead of the

organic carbon content. This yields an organic-matter-partition coefficient, Kqn,

Kom = (Kp * 100)/7oOM (3)

Several regression equations, developed with physicochemical properties such as the n-

octanovwater partition coeffrrcient (Ko*) or aqueous solubility have been used to estimate K6ç

or Ko,n.l3-18 Topological descriptors (e.g. molecular connectivity indices) have also been

used to predict K6ç or Ko-.I, 19-23 Hsvlsver, many of these correlations were developed for

specific cliasses of compounds (e.g. PAFII3) and are not applicable to particular chemicals such

as acids and bases. Furthermorc, some of these models fail by using in the same data set Ks6

and Ksc values without correction factors. Indeed, the relationship K6ç = 1.724 * Kotl must

be used to convert Kom to Ko..l, 15,23 !xs1, in ecotoxicology modeling, it is obvious that Kp

is more inæresting to model than K66 or Ksc. Under these conditions, the aim of this study

was to prcsent a Kp equation easily incoqporable in the main environmental fate models and

allowing the prediction of the sorption behavior of structurally diverse chemicals in various

sediments and soils.

MODEL DEVELOPMENT

Kp values for 53 organic chemicals including aliphatics, aromatics, pesticides, PCB, PAH, and

related compounds (Table 1) were retrieved from the literature.9, 13' 16' 24-34 fhsy $'g1s

selected according to the following criteria:- only experimental results obtained from Freundlich isotherms with N = I werc selected in

order to standardize the units,35- Kn data recorded in soils or sediments having a goOC < 0.1 were rejected since for organic-

poor sorbents, the interactions of the chemicals with the inorganic matrix of sorbent may

become important,S, 28, 36, 37- values conesponding to suspended sediments were rejecæd.When necessary, the relationship ToOC = 7oOMll.724 was used to convert VoOM to VoOC.The log Kqu,, and pKa values (fàUte l) were obtained from the literature.lS,l6'24,2G30,32,38-49 Blpsrimental values or data obtained from critical compilations were preferentially

selected.

33

rt73

Table 1: Training set.

Ctremical log Ko* PKA ToOC pH

AcetophenoneAcridineAcrylonitrile2-Aminoanthracene6-AminochrysenefuiisoleAnthracene-9-carboxYlic acidBenzamideBenzene

n-ButylbenzeneCarbon tetrachlorideCtrlorobenzene

2-ChlorobiphenylChloroform3-Chlorophenol1,2:5,6-Dibenzanthracene1,2-Dichlorobenzene

1,3-Dichlorobenzene1,4-Dichlorobenzene

2,2' -Dichlorobiphenyl

2,4'-Dichlorobiphenyl1,2-Dichloroethane3,4-Dichlorophenol7, I 2-Dimettrylbenzanthracene1,4-Dimethylbenzene

2,4-Dinitro-o-cresolEthylbenzeneFuranHexachlorobenzeneHexanoic acidMethoxychlor

1.59Q4)3.40(38) 5.63(3e)0.12Q6)4J3@) 4.10G8)4.99Q7) 3.70(38)2.1 l (16)4.39(38) 3.6Stlsl0.64(38)2.1,:4J6)23(J6)4.13(28)2.64Q6)2.8406)2.84Q6)2.84Q6)4.51(16)1.97Q6)2.S0Qe) 8.85(40)6.50(30)3.38(16)3.33(t6)3.38(16)3.38(16)3.39(16)3.39(16)4.80(16)5.10(16)1.45(41)3.44(2) 7 39(40,5.98(30)3.15(28)3.15(28)2.85(43) 4.4692',)3.15(16)134Q6'5.7394)1.90(45) 4.85(46)5.08(13)

0.rs-2.38 4.54-8.32t.250.66-1.49r .100.150.66-1.490.66-r.491.100.151.100.66-r.492.80 8.000.11-2.380.66-r.49l . l00.93l . l01 .100.151.101.100.932.80 6.00-8.000.15-2.380.66-1.490.150.15-3.04 4.n-8.291.100.66-1.490.664.85 2.800.13-3.29

l2Q4*)tz?s)2Q6)B@)B@)l(16)l2Q7)1(e)2Q6)l(16)l(28)2Q6)2Q6)t (16)l(28)1(16)2Q6)t(ze)l4(30)2Q6)t (16)l (31)1(16)l (16)t(28)l(16)l (16)l (31)2Qe')l3(30)2Q6)t(28)l2Q2)l(16)2Q6)t(26)t(e)l3(13)

0.15-238**0.48-2.380.66-1.490.r5-2.380.15-2.38l . t0

4.54-7.93**

4.54-8.324.54-8.32

34

tt74

Table I (continued)

Chemical n# log Kes, pKa ToOC pH

3-MethylcholanthrcneNitrobenzenePentachlorophenolPyrene

Silvex2,3,7,8-TCDDI,2,3,4 -T elrachlorobe nzene1,2,4,5 -T etrachl orobe nzene1,1,2,2-T etrachlo roethaneTetrachloroethylene2,3,4,6 -T etrach lo rophenolTetrahydrofuran1,2,4,5 -T etramethylbenzeneToluene

1,2,3 -Trichlorobenzene

1,2,4-Trichlorobenzene

2,4,4' -T rtchlo robiphenyl1 , 1 ,l -Trichloroethane

2,4,5 -Trichlorophenol

1,2,3 -Trichloropropâne

1,2,3 -Trimethylbenzene

1, 3,5 -Trimethylbenzene

6.42Q0)t.87Q6)5.04(2) 4.92(|,0)5.99(30)5.09(30)3.4l@7) 3.07g2',)6.42@8)4.64(u)4.72Q8)239Qr)2.ffiQ8)4.42((2) 538(40)9.46Q6)4.05(28)2.71Q6)2.71Q6)4.14(M)4.14(M\4.0206)4.0206)5.6206)2.47@r)3.72/4e> 7.$@o)2.Ola6)3.60(28)3.60(28)

0.1l-2.38*'r0.66-1.493.20 4.700.11-2.380.r3-3.290.15-3.04 4.27-8.29**0.660.150.150.930.15t.70-3.20 3.404.700.66-r.490.150.66-r.490.150.154.70r .100.15t . l00.932.80 8.000.66-r.490.150.15

l2(30*)2Q6)t(2e)l4(30)tz(r3)t2Q2)1(33)1(28)l(28)1(31)l(28)2Qe)2Q6)l(2E)2Q6)l(28)I (28)

l(34)l(16)t(28)l (16)l (31)t(ze)2Q6)l(28)l(28)

#number of experimental Kp values; *reference; **range.

Regression analysis of the 229 Kp values obtained from the 53 chemicals under study versustheir log Ksç, and log fsc values (Table l) yields Eq. 4.

log Kp = 0.96 log Kso, + 1.04 log fsç - 0-13 (4)

n=229 s = 0.639 r = 0.924 F=660.07 p

I 175

opposite, Fig.distributed.

Frequencies

I reveals that the residuals for the other chemicals under study are well

El AcidsI Bases@ Other

compounds

Residuals

Fig. 1. Graphical analysis of the residuals (log Kp obs. - log Kp calc.) obtained with Eq. 4.

Our results are not surprising since numerous autho632, 36, 39' 50-53 have underlined that it

was dangerous to neglect the role of pH in the modeling of the adsorption of acids and bases

in soils and sediments. Indeed, it is well admined that most weakly acidic chemicals are in

predominantly ionic (i.e. negatively charged) form at the pH of most natural soils, while most

weakly basic chemicals are in molecular form. In either case, as natural sorbent pH decreases

toward a value equal to the dissociation constant (pKa) of the chemical, sorption tends to

increase because at low pH, the acidic chemical has more of the molecular species and the

basic chemical has more of the protonated (positive) ionic species.S The colloidal surfaces of

most natural soils are negatively charged and therefore have an affinity for positively charged

molecules, but not much affinity for negatively charged mslssulss.S' 54

Therefore, to optimize the model, it could be interesting to introduce two different corection

factors (one for acids and another for bases) allowing to quantify the variations of the

concentntion of ionic species in the range of soil pH values.

Thus, for acids, the evolution of the anionic species concentration in relation to pH values can

be estimated by:55

CFa = logT+TOPH-PKA

For bases, the relation between the protonated species concentration and the pH values can be

expressed by:55

(s)

-1.8 -1.6 -1.4 -1.2 -l -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 L 1-2 l-4

36

t t76

CFb = log

1+ l0pKa-pH

(6)

Introduction of CFa and CFb in Eq. 4 allows to obtain the following model:

log Kp = 0.92log Ko* +

n=229 s=0 .453

1.09log fo. + 0.33 CFa - 1.32 CFb + 0.30 (7)

r = 0.963 F=713.57 pcO.OlVo

Like Wauchope and coworkers,56 we have defined acids as the sole chemicals whose neutral(molecular) forms are capable of forming negatively charged ions. Therefore, in Eq. 7, CFamust be taken equal to zero for bases and other compounds since this parameter is null whenthe concentration of anionic species is zero.55 In the same way,56 rve have defined bases as thesole chemicals whose neutral forms are able to form positively charged ions. Therefore inF;q.7, CFb must be taken equal to zero for acids and other compounds since this parameter isnull when the concentration of cationic species is zero.55Eq.l gives a better distribution of the residuals for the acids and bases (Fig. 2) than Eq. 4(Fig. l), but some Kp values for basic chemicals are still underestimated.

Frequencies E Acidst Bases@ Other

compounds

Residuals

I

-1.8 -1.6 -r.4 -r.2 -l -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 | r.2 t.4

Fig.2. Graphical analysis of the residuals (log Kp obs. - log Kp cab.) obtained with Eq. 7.

The remaining difference between the adsorption behavior of bases and acids could be

37

tt77

explained by the fact that the adsorption of basic compounds in certain soils (e.g.

montmorillonite clay systems) is principally dependent upon the surface acidity and not upon

the pH of the bulk solution, while the converse is true for the adsorption of acidic

compounds.50 Bailey et al.So have estimated that the surface acidity of montmorillonite

upp"àr, to be 34 pH units lower than the pH of bulk solution. Green and KarickhoffS have

stressed that a reasonable estimate of surface acidity (pHs) is two pH units lower than the bulk

suspension pH (i.e. pHs = pH - 2). Under these conditions, we have tried to introduce this

conection in our model (Eq. 7) defining the new following variable:

CFb' = logl+10pKa- (pH-2)

Introduction of CFb' in Eq. 7 instead of CFb yields Eq' 9'

log Kp = 0.93 log Ks,r, + 1.09 log fqc + 0.32 CFa - 0.55 CFb' + 0'25 (9)

n=229 s=0.433 r = 0.966 F=786.07 p

l l78

Introduction of CFb' in the modelcompounds (Fig.4).

Frequencies

leads to a better distribution of the residuals of basic

E AcidsI BasesE other

compounds

Residuals

-1.8 -1.6 -t.4 -r.2 -l 4.E -0.6 -O.4 -0.2 0 0.2 0.4 0.6 0.8 | r.2 t.4

Fig. 4. Graphical analysis of the residuals (log Kp obs. - log Kp carc.) obtained with Eq. 9.

MODEL VALIDATION

To estimate the predictive power of our final model (Eq. 9), a test set of 500 experimental Kpvalues for 87 chemicals has been constituted (Table 2). It is important to note that these valuesdid not satisfy all the constraints defined for the selection of the original raining set sincethey were generally obtained from Freundlich isotherms with 0.8 < N < 1.2 or by one initialCgncentratiOn.9, 15, 29, 34, 39, 51, 57 -7 6The log Ksç, and pKa values (Table 2) were obtained from the literature.4, 15, 16,29,38-46,49,52,59,60, 63,70,71,75-85 Experimental values or data obtained from critical compilationswere preferentially selected.

Table 2: Test set.

IE

Chemical log Ksyy pKa N ToOC pH

AcetanilideAcridineAlachlor

2(9")l(3e)l(s7)

1.16(38) 0.82-0.84t 1.58-4.85t3.40(38) 5.68(3e) 1.02 0.58 7.642.64Q7',) l.0l t.t1

39

tt79

Table 2 (continued)

log K6rry pKa

AldicarbAldrinAmetryneAnilineAtrazine

BenzamideBenzoic acid (B.a.)

B.a. ethyl esterB.a. methyl esterB.a. phenyl ester4-BromophenolCaptafolCaptanCarbendazimCarbarylCarbofuranChlorfenvinphosChlorobenzene2-Chlorophenol3-Chlorophenol

Chlorsulfuron

3-Cresol2,4-D

DiazinonDibenzottriophene2,4-Dichloroaniline3,5-Dichloroaniline1,4-Dichlorobenzene2,4-Dichlorophenol3,4-Dichlorophenol

Dieldrin

1(15*)1(1s)34(5E)l(se)4(60)l(61)25(62)1(e)l(e)3(el3(e)Z(ell (15)l (15)l(ls)l5(63)1(15)2(64)1(1s)1(6s)1(66)5Qe)3(34)1(66)1(67)Z3(68)1(66)3(e)19(6e)1(1s)1 1(70)3Q4)1(5e)1(6s)3(71)s?e)4Q4)l(1s)

1.08(77)5.66(4)2.58(78)0.90(38)233Q7)233Q7)233Q7)0.64(38)1.87(38)2.64G8)z.z0Q8)3.59(38)2.590s)3.83(15),.540s)1.5263)23205)t.63Q7)3.19(ls)2.84(16)z. t7@3)z.s}Qe)2.50Qe)2.50Qe)2.20(80)2.20(80)1.98(38)2 .81( t l )2 .81(81)3. I l ( ls)4.38(70)2.78$2)2.69( 5e)

3.39(16)3.21(2)3.44@2)3.44@2)4.60(4)

5.703.40-6.003.60-5.905.704.ffi5.20-7.n5.702.80-7.r06.30-8.30

t$t$ 2.Os

4.gg(7e) 1$ 0.35-20.88t4.50-9.0074.60(46) 0.83 0.849 5.40$1.68(7e) I 0.40-7.60 4.40-8.00r.689er 0.92 0.43 6.051.63(7e) 1$ 0.35-2.84 4.30-7.10

0.88 4.854.20(46) 0.90 4.85 2.80

0.81-0.88t 1.25-4.850.81-0.85 1.25-4.850.91-0.93 1.25-1.58

9.34(46) 1$ 1.46 6.70t$ Z.0St $ 2.Os

4.43(63) l$ 0.63-3.47 4.14-7.54r $ Z.Os0.90-0.94 1.60-2.50t$ 2.050.87 4.70

8.49(40) 0.80 2.968.85(40) 0.80 r.70-3.208.85(40) 0.85-0.96 2.15-9.058.85(40) 0.83 2.963.60(52) 0.87 1.423.69(s2) 1$ 0.95-20.3610.0(46) 0.87 2.962.64@6) 0.89-1.06 1.25-4.852.64@6) 0.82-l .16 0.28-2.73

r$ 2.Os0.88-1.10 0.15-2.38

2.05(83) 0.85-1.11 3.54-9.05237(46) 0.91 0.84$

0.96 2.157.63(a0) t$ 0.20-3.707 3g(.J0') 0.80-0.90 0.90-3.207 39@o) 0.81-0.94 2.15-9.05

l$ 2.05

3.60-s.905.40$

4.20-7.403.40-7.503.60-5.90

ll()

l t80

Table 2 (continued)

log Ksw pKa pH

DimethoateNIV-DimethylanilineDiuronFenamiphosFolpetHexachlorobenzeneHexanoic acid3-MethoxyphenolN-Methylaniline4-Methylaniline4-Mettrylbenzoic acidMetsulfuron-methyl4-Nitroaniline4-Nitrobenzoic acid2-NitrophenolParathion

2,2"5,5'-PCB

3,3"4,4'-PCB212"3,4,s',-PCB2,2"3,5"6-PCB2,2"3"415-PCB2,2"4,5rs'.-PCB

2,3,3"4"6-PCB2r3"4r4"5-PCB2,2"3,4,4"5-PCB2,2"3,4,5,''-PCB212"314"5"6-PCB2,2"4,4"5,5'-PCB

PentachlorobenzenePentachlorophenol

PhenolPhenylacetic acid

l (15*)3(e)34(58)l ( ls)t(ls)l (15)2Q)l(66)l(e)l(e)t(e)2368)1(e)1(e)l(66)1(15)6Q2)2Q3',)5(74)5Q4)2Q3)2Q3)2Q3)2Qt>5(74)2Q3)2Q3\2Q3)2Q3)2Q3)2Qt>5Q+'.12Q3'.)6Qe)4Q4>3Qr>l(ls)2(e)

0.79(ls)2.31(38)2.60Q7)3.18( ls)3.63(ls)5.73(u)1.90(45)1.58(4e)1.66(38)1 .39(38)234|38',)2.20$0)1.39(4s)1.89(45)1.79@s)3.76Q7)3.76Q7)6.10(84)6.10(84)6.10(E4)6.50(84)6.40(E4)6.60(84)6.40(84)6.40(84)6.53(85)6.40(84)7.00(84)7.59(85)6.80(84)6.90(84)6.90(84)4.949r)5.04G2)5.04142)5.04Q2)1.46(15)l .4 l (38)

s.702.802.802.80

6.702.805.70

3.40-7.503.60-5.904.20-7.406.701.254.85

t $ 2.055.15(46) 0.87-0.91t 1.25-4.85T 2.80-7.10f

r $ 0.35-20.881$ 2.05r$ 2.05I $ 1.53

4.35(46) l.0l 1.25-1.58 6.70-7.r0

3.36(80) t$ 0.95-20.36 5.20-7.90

9.65(46) 0.89 2.964.85(46) 0.89 4.855.08(38) 0.96 4.85436(46) 0.82 4.85

l .0 l (46) 0.81 1.253.44@6) l.r4 4.857.22Q6) 0.89 2.96

t$ 2.051.02-l. l | 0.44-14.28t $ 0.16-1.87r $ 0.16-1.87I $ 0.16-1.87r $ 0.16-1.87I $ 0.16-1.87r $ 0.16-1.87r $ 0.16-1.871$ 0.16-1.87l $ 0.16-1.87t $ 0.16-1.87r $ 0.16-1.87t $ 0.16-1.87l$ 0.16-1.87l $ 0.16-1.87r $ 0.16-1.871$ 0.16-1.87

4.92G0) 0.80-0.90 0.90-29.804.92/40) 0.82-0.92 2.15-9.054.92(a0) l$ 0.20_3.709.99(46) 1$ 1.464.3 t@6) O.95-r.t2 2.80-6.70

4L

1 181

Table 2 (continued)

log Ke'r,, pKa

PhoratePrometone

PrometrynePropazinePyridine

Quinoline

Simazine

2,4,5-T

L,2,3,4-TCBg

1,2,3,s:lc,BfTetrachloroguaiacol2,3,4,'-TCP&2,3,4,6-TCP&

3,4,5-Trichloroaniline 6Q6)1,2,3-Trichlorobenzene 3(34)

2Q3)1,2,4 -T richlorobenzen e 2Q 3)

1,3,5-Trichlorobenzen "

2Q 3)

4,5,6-Trichloroguaiacol 3(7 I )

2,4,5-TichloroPhenol 8(29)4Q4)

2,4,1-TichloroPhenol 3Ql)Trifluralin 4(60)

3.93Q7) 1$ 2.051.9 4Q 8) 4.28Q e) 0.92-0.991 0.33-0.43tt.g4QB) 4.2gQe) l$ 0.35-2.842.gg(78) 4.95(7e) l$ 0.35-2.843.09(78) 1.35(7e) l$ 0.35-2.840.65(38) 5.23Qe) t.04 0.58

2.03(38) 4.92Qe) 0.87 0.582.03(38) 4.92Ge)0.84-0.91 0.35-0.582.27Q8) 1.65(7e) l$ 2.052.27Q8) 1.65(7e) 1$ 0.35-2.84

3.36(8r) 2.90Qs) 0.84-1.14 1.25-4.85

3.36(8r) 2.90Qs) 0.81-0.85 0.46-1.744.64(M) 0.91-1 .13 2.15-4.704.64@) 1$ 0.16-1.874.469D I $ 0.16-1.874.45@2) 5.97(71) 1$ 0.20-3.104.82Q2) 6.96(40) 0.81-0.82 2.15-3.54

4.4292) 5.39(40) 0.80-0.90 0.90-29.80

4.42/12) 5.39(a0) t$ 0.20-3.70

3.49Q6) 1.78(83) I 0.12-6.344.r4(M) 0.96-1.01 2.15-9.05

4.14(M) I $ 0.16-1.874.0206) 1$ 0.16-1.874.rg(M) t$ 0.16-1.873.74@2) 7.49(71) l$ 0.20-3.70

3.72!4e) 7.43(40) 0.80-0.90 0.90-29.80

3.7z3e\ 7 .$@0\ 0.81-0.98 2.r5-9.0s3.75(42) 7.42(J0) tS 0.20-3.70

5.07(60) I 0.40-7.60

1(15*)2(6r)2562)2562)25(62)t(3e)1(3e)2(5r)1(15)25(62)

3(e)2Qs)2(6s)2Q3)2Q3)3(71)2Q4)s?e)3(71)

6.05-6.3074.30-7.r04.30-7.r04.30-7.107.&7.&7.46-7.&6.104.30-7.102.80-7.105.90-7.70

4.20-7.404.80-5.604.60-7.504.20-7.404.50-5.10

4.20-7.403.40-7.503.60-5.904.20-7.40

# number of experimental Kp values*reference; Trangei $single concentrat ion; $ref . 86; tTetrachlorobenzene;

&Tetrachlorophenol.

Comparison of Figs. 5 to 7 shows that Eq. 9 is the most suitable model to describe the

sorption behavior of the 87 organic chemicals listed in Table 2.

42

tt82

log Kp calc.6

5

4

3

2

I

0

- l

-2 log Kp obs.

Fig. 5. Observed versus calculated (Eq. 4) Kp values for the test set.

log Kp calc.

6

5

4

3

2

I

0

- l

-2 log Kp obs.- 2 -1 0123456

Fig. 6. Observed ve$us calculated (Eq. 7) Kp values for the test set.

- 2 -1 0 r23456

43

l 183

log Kp obs.

Fig. 7. Observed versus calculated (Eq. 9) Kp values for the test set.

CONCLUSION

Our study shows that it is possible to propose a unique general QSAR equation (Eq. 9) for

estimating the sorption behavior of all the organic chemicals which can potentially

contaminate the ecosystems. Introduction of Eq. 9 in the environmental fate models (e.g.

fugacity models6, 87) could be useful to increase the accuracy of their outputs and stretch

their domain of application. It could be also interesting to test the usefulness of our correction

factors for ionization problems to improve the predictive power of the models based upon

linear relationships between soil and/or sediment sorption coefficients and topological

descriptors such as molecular connectivity indices.

ACKNOWLEDGEMENTS

V/e acknowledge with appreciation the comments of Pr. M. Chastrette (Laboratoire de

Ctrimie Organique Physique, University Lyon I, France) and Dr. T. E. McKone (Health and

Ecological Assessments Division, Lawrcnce Livermore National Laboratory, California)which improved this paper.

4

l l 84

REFERENCES

l. Bahnick, D.A. and Doucette, WJ. (1988). Use of molecular connectivity indices to estimate

soil sorption coefficients for organic chemicals. Chemosphere 17,1703-1715.

2. Domine, D., Devillers, J., Chastrette, M., and Karcher, W. (1992). Multivariate strucftre-

properry relationships (MSPR) of pesticides. Pestic. sci. 35, 73-82.

3. Devillers, J., Karcher, W., and Isnard, P. (1991). Graphical display on the frrgacity model

level I. Chemosphere 23, 633-642.4. Devillers, J., Thioulouse, J., Domine, D., Chastrette, M., and Karcher, W' (1991)'

Multivariate analysis of the input and output data in the fugacity model level I. ln, Applied

Multivariate Analysis in SAR and Environmental Studies (J. Devillen and W. Karcher, Eds.).

Kluwer Academic Publishers, Dordrecht, pp. 281-345.

5. Jaffé, R. (1991). Fate of hydrophobic organic pollutants in the aquatic environment: a

review. Environ. Pollut. 69, 237 -257 .

6. Mackay, D. (1991). Multimedia Environmental Models. The Fugaciry Approach. Lewis

Publishers, Chelsea, P. 257 .7. McKone, T.E. (1993). The precision of QSAR methods for estimating intermedia transfer

factors in exposure assessments. SAR QSAR Environ. Res. 1, 4l-51.

8. Green, R.E. and Karickhoff, S.W. (1990). Sorption estimates for modeling. In, Pesticides

in the Soil Environment: Processes,lmpacts, and Modeling (H.H. Cheng, Ed.). Soil Science

Society of America, Madison, pp. 79-l0l-9. von Oepen, B., Kôrdel, W., and Klein, W. (1991). Sorption of nonpolar and polar

compounds to soils: processes, measurements and experience with the applicability of the

modified OECD-Guideline 106. C hemo sphere 22, 285 -304'

10. Lambert, S.M. (1968). Omega (O), a useful index of soil sorption equilibria. J. Agric.

Food Chem. L6,340-343.ll. Grathwohl, P. (1990). Influence of organic matterfrom soils and sediments from various

origins on the sorption of some chlorinated aliphatic hydrocarbons: implications on Keç

correlation s. Environ. Sci. T echnol. 24, 1687 -l 693.

12. Rutherford, D.W., Chiou, C.T., and Kile, D.E. (1992).Influence of soil organic matter

composition on the partition of organic compounds . Environ. Sci. Technol.26,336-340.

13. Karickhoff, S.W., Brown, D.S., and Scott, T.A. (1979). Sorption of hydrophobic

pollutants on natural sediments. Wat. Res.13,241-248.

14. Kenaga, E.E. and Goring, C.A.I. (1980). Relationship between water solubility, soil

sorption, octanol-water partitioning, and concentration of chemicals in biota. ln, Aquatic

Toiicotogy (J.G. Eaton, P.R. Parrish, and A.C. Hendricks, Eds.). ASTM STP 707, pp' 78-

115 .15. Briggs, G.G. (1981). Theoretical and experimental relationships between soil adsorption,

octanol-water partition coefficients, water solubilities, bioconcentration factors, and the

parachor. J. Agric. Food Chem.29, 1050-1059.16. Ctriou, C.T., Porter, P.E., and Schmedding, D.W. (1983). Partition equilibria of nonionic

organic compounds between soil organic matter and water. Environ. Sci. Technol.17,227'

45

r l8s

23r.17. Chin, y.p., peven, C.S., and Weber, W.J. (1988). Estimating soiVsediment partition

coefficients for organic compounds by high performance reverse phase liquid

chromatography. Wat. Res. 22, 87 3-881.

18. Noegrohati, S. and Hammers, w.E. (1992). Regression models for some solute

distribution equilibria in the terrestrial environment. Toxicol. Environ. Chem- 34' 175-185'

1g. sabljic, A. and protic, M. (1982). Relationship between molecular connectivity indices

and soil sorption coefficients of polycyclic aromatic hydrocarbons. Bull. Environ. Contam.

Toxicol.2E, 162'165.20. Koch, R. (19g3). Molecular connectivity index for assessing ecotoxicological behaviour of

organic compounds . Toxicol. Environ. Chem. 6,87-96'

Zt. SaUtiic, A. (1984). Predictions of the nature and strength of soil sorption of organic

pollutants by molecular topology. J. Agric. Food Chem.32,243-246.

)2. S"Ut;i., A. (1987). On rhe prediction of soil sorption coefficients of organic pollutants

from molecular structure: application of molecular topology model. Environ. Sci. Technol'

21,358-366.23. Meylan, W., Howard, P.H., and

contribution method for predicting

1560-1567.24. Khan, A., Hassett, J.J., Banwart, W.L., Means, J.C., and Wood, S.G. (1979). Sorption of

acetophenone by sediments and soils. soi/ sci. 12E,297-302.

25. Banwart, W.L., Hassett, J.J., Wood, S.G., and Means, J.C. (1982). Sorption of nitrogen-

heterocyclic compounds by soils and sediments. Soi/ Sci. 133, 4247.

26. Walton, B.T., Hendricks, M.S., Anderson, T.A., Griest, w.H., Merriweather, R',

Beauchamp, J.J., and Francis, C.W. (1992). Soit sorption of volatile and semivolatile organic

compounds in a mixture. J. Environ-Qu.a\.21,552-558'27. Means, J.C., Wood, S.G., Hassett, J.J., and Banwart, W.L. (1982). Sorption of amino- and

carboxy-substituted polynuclear aromatic hydrocarbons by sediments and soils. Environ. Sci.

Technol. l6, 93-98.2g. Schwarzenbach, R.P. and Westall, J. (1981). Transport of nonpolar organic compounds

from surface water to groundwater. Laboratory sorption studies. Environ. Sci. Technol'15,

1360-1367.29. Lagas,P. (1988). Sorption of chlorophenols in the soil. Chemosphere 17,205-216'

30. Means, J.C., 'Wood, S.G., Hassett, J.J., and Banwart, W.L. (1980). Sorption of

polynuclear aromatic hydrocarbons by sediments and soils. Environ. Sci. Technol.14, 1524'

1528.31. Chiou, C.T., Peters, L.J., and Freed, V.H. (1979). A physical concept of soil-water

equilibria for nonionic organic compounds . science 206,831'832.

32. Jafvert, C.T. (1990). Sorption of organic acid compounds to sediments: initial model

developme nt. Env iron. T oxicol. C hem. 9, 1259 -1268'

33. Walters, R.W., Ostazeski, S.A., and Guiseppi-Elie, A. (1989). Sorption of 2,3,7,8-

tetrachlorodibenzo-p-dioxin from water by surface soils. Enviro n. Sci. Technol.z}, 480'484,

Boethling, R.S. (1992). Molecular topologyÆragment

soil sorption coefficients. Environ. Sci. Technol'26,

6

1186

34. van Gestel, C.A.M. and Ma, W.C. (1990). An approach to quantitative structure-activityrelationships (QSARs) in earthworm toxicity studies. Chemosphere2l,1023-1033.35. Bowman, B.T. (1982). Conversion of Freundlich adsorption K values to the mole fractionformat and the use of Sy values to express relative adsorption of pesticides. SoiI Sci. Soc. Am.J. 46,740-743.36. Schellenberg, K., Leuenberger, C., and Schwarzenbach, R.P. (1984). Sorption ofchlorinated phenols by natural sediments and aquifer materials. Environ. Sci. Technol.lt,652-657.37. Karickhoff, S.W. (1984). Organic pollutant sorption in aquatic systems. J. Hydraul. Eng.110,707-735.38. Sangster, J. (1989). Octanol-water partition coefficients of simple organic compounds. ./.Phys. Chem. Ref. Data lE, ll ll-1229.39.7-achara, J.M., Airtsworth, C.C., Cowan, C.E., and Thomas, B.L. (1987). Sorption ofbinary mixtures of aromatic nitrogen heterocyclic compounds on subsurfarce rnaterials.Environ. Sci. Technol.2l, 397 -402.

40. Jones, P.A. (1981). Chlorophenols and their Impurities in the Canadian Environment.Environment Canada, Report SPE 3-EC-81-2F, p.322.41. Veith, G.D., Macek, K.J., Petrocelli, S.R., and Carroll, J. (1980). An evaluation of usingpartition coefficients and water solubility to estimate bioconcentration factors for organicchemicals in fish. ln, Aquatic Toxicology (J.G. Eaton, P.R. Panish, and A.C. Hendricks,Eds.). ASTM STP 707, pp.116-129.42. Xie, T.M., Hulthe, B., and Folestad, S. (1984). Determination of partition coefficients ofchlorinated phenols, guaiacols, and catechols by shake-flask GC and HPLC. ChemosplurelS,M5-459.43. Westall, J.C., Leuenberger, C., and Schwanenbach, R.P. (1985). Influence of pH andionic strength on the aqueous-nonaqueous distribution of chlorinated phenols. Environ. Sci.Technol. 19,193-198.44. de Bruijn, J., Busser, F., Seinen, W., and Hermens, J. (1989). Determination ofoctanoVwater partition coefficients for hydrophobic organic chemicals with the "slow-stirring" method. Environ. Toxicol. Chem.8, 499-512.

' 45. El Tayar, N., Testa, B., and Camrpt, P.A. (1992). Polar intermolecular interactionsencoded in partition coefficients: an indirect estimation of hydrogen-bond parameters ofpolyfunctional solutes. J. Phys. Chem.96, 1455-1459.46. Dean, J.A. (1985). Lange's Handbook of Chemistry, thirteenth edition. McGraw-Hill

I Book Company, New York.47. Howard, P.H. (1990). Handbook of Environmental Fate and Eryosure Data for OrganicChemicals ///. I-ewis publishers, Chelsea, p. 684.

, 48. Sijm, D.T.FI.M.,'Wever, [tr., de Vries, P.J., and Opperhuizen, A. (1989). Octan-l-ofwaterpartition coefficients of polychlorinated dibenzo-p-dioxins and dibenzofurans: experimentalvalues determined with a stining method. Chemasphere 19,263-266.49. Leo, A., Hansch, C., and Elkins, D. (1971). Partition coefficients and their uses. Chem.Rev.71,525-616.

47

1187

50. Bailey, G.W., White, J.L., and Rothberg, T. (1968). Adsorption of organic heôicides bymontnorillonite: role of pH and chemical character of adsorbate. Soil Sci. .Soc. Amer. Proc.32,222-234.5l.7achara, J.M., Ainsworth, C.C., Felice, L.J., and Resch, C.T. (1986). Quinoline sorptionto subsurface materials: role of pH and retention of the organic cation. Environ. Sci. Technol.20, 620-627.52. Nicholls, P.[I. and Evans, A.A. (1991). Sorption of ionisable organic compounds by fieldsoils. Part l: acids. Pestic. Sci.33, 319-330.53. Nicholls, P.H. and Evans, A.A. (1991). Sorption of ionisable organic compounds by fieldsoils. Part 2: cations, bases and zwitterions. Pestic. Sci. 33, 331'345.54. Green, R.E. (1974). Pesticide-clay-water interactions.ln, Pesticides in Soil and Water(W.D. Guenzi, Ed.). Soil Science Society of America, Madison, pp.3-37.

55. Scherrer, R.A. and Howard, S.M. (1977). Use of distribution coefficients in quantitative

structurc-activity rclationships. l. Med. Chem. 20, 53-58.56. Wauchope, R.D., Buttler, T.M., Homsby, 4.G., Augustijn-Beckers, P.W.M., and Burt,

J.P. (1992). The SCS/ARS/CES pesticide properties database for environmental decision-making. Rev. Environ. Contam. Toxicol.123, 1-155.57. Sato, T., Kohnosu, S., and llartwig, J.F. (1987). Adsorption of butachlor to soils. n/.Agric. Food Chem. 35, 397 -402.

58. Liu, L.C., Cibes-Viadé, H., and Koo, F.K.S. (1970). Adsorption of ametryne and diuronby soils. Weed,Scf. 18, 470-474.59. Dao, T.H., Bouchard, D., Mattice, J., and k"y, T.L. (1986). Soil sorption of aniline and

chloroanilines: direct and indirect concentration measurements of the adsorbed phase. Soil

Sci. 141, 26-30.60. Francioso, O., Bak, E., Rossi, N., and Sequi, P. (1992). Sorption of atrazine andtrifluralin in relation to the physico-chemical characteristics of selected soils. Sci. TotalEnviron. 123 I L24, 503-512.61. Singh, G., Spencer, W.F., Cliath, M.M., and van Genuchten, M.Th. (1990). Sorptionbehavior of s-triazine and thiocarbamate herbicides on soils. J. Environ.Qwl.19,520'525.62. Talbert, R.E. and Fletchall, O.H. (1965). The adsorption of some s-triazines in soils.Weeds 13,46-52.63. Austin, D.J. and Briggs, G.G. (1976). A new extraction method for benomyl residues insoil and its application in movement and persistence studies. Pestic. Sci.7,2Ol-210.64. Achik, J., Schiavotr, M., and Jamet, P. (1991). Study of carbofuran movement in soilspart I-soil structure. Environ. Int.17,73-79.65. van Gestel, C.A.M., Ma, W.C., and Smit, C.E. (1991). Development of QSARs interrestrial ecotoxicology: earthworm toxicity and soil sorption of chlorophenols,chlorobenzenes and dichloroaniline. Sci. Total Environ. 109 I ll0, 589-604.66. Boyd, S.A. (1982). Adsorption of substituted phenols by soil. Soif Sci. 134,337-343.67. Menie, \M. and Foy, C.L. (1986). Adsorption, desorption, and mobility of chlorsulfuronin soils. J. Agric. Food Chem.34,89-92.68. Walker, A., Cotterill, E.G., and lVelch, S.J. (1989). Adsorption and degradation of

48

I 188

chlorsulfuron and metsulfuron-methyl in soils from different depths. Weed Res.29,281-287.69. Hermosin, M.C. and Cornejo, J. (1991). Soil adsorption of 2,4-D as affected by the claymineralogy. Toxicol. Environ. Chem. 3l-32, 69-77.70. Hassett, J.J., Means, J.C., Banwart, W.L., Wood, S.G., Ali, S., and Khan, A. (1980).

Sorption of dibenzothiophene by soils and sediments../. Environ. Qual.9, 184-186.

7L. Seip, H.M., Alstad, J., Carlberg, G.E., Martinsen, K., and Skaane, R. (1986).

Measurement of mobility of organic compounds in soils. Sci. Total Environ.s0, 87-101.

72. Wahid, P.A. and Sethunathan, N. (1978). Sorption-desorption of parathion in soils. J.

Agric. Food Chem.26, 101-105.73. Payâ-Pêrez, A.8., Riaz, M., and Larsen, B.R. (1991). Soil sorption of 20 PCB congenersand six chlorobenzeîes. Ecotoxicol. Env. Safery 21, l-17.74. Cortes, A., Riego, J., Payâ-Pérez, A.B., and Larsen, B.R. (1991). Soil sorption of co-planar and non-planar PCBs. Toxicol. Environ. Chem.Sl'32,79-86.75. Koskinen, W.C, O'Connor, G.A., and Cheng, H.H. (1979). Characterization of hysteresis

in the desorption of 2,4,5-T from soils. Soi/ Sci. Soc. Am. J.43,871-874.76. Payâ-Pêrez, A.B. and Pelusio, F. (1992). Sorption of 3,4,5-trichloroaniline in a soilprofîle. Sci. Total Environ. 1231124, 519-529.77. Suntio, L.R., Shiu, W.Y., Mackay, D., Seiber, J.N., and Glotfelty, D. (1988). Critical

review of Henry's law constants for pesticides. Rev. Environ. Contam. Toxicol.l03, 1-59.

78. Gerstl, Z. and Helling, C.S. (1987). Evaluation of molecular connectivity as a predictive

method for the adsorption of pesticides by soils. Environ. Sci. HealthB2z, 55-69.79. Weber, tB. (1970). Mechanisms of adsorption of s-triazines by clay colloids and factors

affecting plant availability. Re sidue. Rev. 32, 93- I 30.80. Grayson, B.T. and Kleier, D.A. (1990). Phloem mobility of xenobiotics. IV. Modelling of

pesticide movement in plants. Pestic. Sci. 30, 67-79.81. Garten, C.T. and Trabalka J.R. (1983). Evaluation of models for predicting tenestrialfood chain behavior of xenobiotics. Environ. Sci. Technol. ll, 590-595.82. Hulzebos, E.M., Adema, D.M.M., Dirven-van Breemen, E.M., Henzen, L., and van

Gesrel, C.A.M. (1991). QSARs in phytotoxicity. Sci. Total Environ.l09l1l0,493497.83. Perrin, D.D., Dempsey, 8., and Serjeant, E.P. (1981). pKaPrediction for Organic Acids

and, Bases. Chapman and Hall, London, p.146.84. Shiu, lV.Y. and Mackay, D. (1986). A critical review of aqueous solubilities, vaporpressures, Henry's law constants, and octanol-water partition coefficients of thepolychlorinated biphenyls. J. Phys. Chem. Ref. Data 15, 9l l'929.85. Sklarew, D.S. and Girvin, D.C. (1987). Attenuation of polychlorinated biphenyls in soils.

Rev. Environ. Contam. Toxicol. 9E, I -41.

r 86. Dao, T.H., Marx, D.8., Lavy, T.L., and Dragun, J. (1982). Effect, and statisticalevaluation, of soil sterilization on aniline and diuron adsorption isotherms. Soil Sci. Soc. Am.

J.46,963-969.i 37. Devillers, J., Bintein, S., Dominê, D., and Karcher,

'W. CHEMFRANCE: A regional

fugacity level III model applied to France. (In preparation).

49

ARTICLE III

50

@ Pcrgamon

Chenosphcn, VoL A[ No. 3, pp. 457476' 1995Copyn$t O 1995 Escvicr Sctcncc Ltd

Èintsd in GtÉt Britrir AU ti8!B rwrvcduI45{535D5 $95m0.æ00456535(94p0425-0

CHEMFRANCE: A REGIONAL LEVEL III FUGACITY MODEL APPLIED TOFRANCE

J. Devillers* and S. Binæin

CTIS, 2l rue de la Bannière,69003 Lyon, France

W. Karcher

Commission of the European Communities, Joint Research Centre, Ispra Establishment,

l-21020 IsPra Varese, ItalY

(Reccivcd in Gcrmany 22 Septcmbcr 1994; rcccpted 3l Octobcr 1994)

ABSTRACT

CHEMFRANCE, a computer model allowing to estimate the environmental fate of organic

chemicals in France, is presented. This multimedia model is represented by six bulk

compartnents (i.e., air, surface water, soil, bottom sediment, groundwater' coastal water),

and by ten subcompargnents. The model employs the fugacity concept and requires

information on chemical properties (i.e., molecular weight, aqueous solubility' vapor

pressurc, n-octanoVwater partition coefficient, dissociation constant, melting point, reaction

half-lives) and emission rates of the pollutants. The ouputs of the model consist of estimated

chemical distribution between environmental media, transport and transformation process

rates, and steady state concentrations in one of the twelve defined regions of France, or

France as a whole, at a chosen season.The nature of the model and its underlying assumptions are described. An illustrative

example dealing with the modeling of the environmental fate of isobutylene is presented.

espects of model validation arc discussed and rccommendations are made for a proper use of

CHEMFRANCE.

* Author to whom all correspondence should be addressed.

457

51

I458

INTRODUCTION

Pollutants discharged into the environment are distributed across environmental media (e.g.,

air, water, soit) and among biota as the result of their physicochemical properties (e.g.,

aqueous solubility, vapor pnessure, n-octanoVwater partition coefficient (log P), dissociation

constant) and the nature of the environment (e.g., climatic parameters).1'2 The human mind

is not able to take into account simultaneously the above parameters in order to estimate the

theoretical distribution of a chemical between the different compartments of the biosphere.

Mathematical models, at different levels of complexity can be used to overcome this problem.

These models are very attractive since they provide a rapid and inexpensive simulation tool to

produce a comprehensive picture of the environmental fate of the organic chemicals.3'4

^A.mong them, the level III fugacity models are particularly powerful to estimate the

environmental distribution of pollutants in abiotic and biotic media.4-6 The assessment is

based on information relative to emission rates, environrnental conditions, and

physicochemical and reactive properties of the organic chemicals.

6 this paper, we describe a level III computer model (CFIEMFRANCE) which can be used to

estimate the environmental fate of organic chemicals in France. The country has been divided

into twelve rcgions where the environment is represented by a multimedia system constinrted

of six bulk compartments (i.e., air, surface water, soil, bottom sediment, groundwater,

coastal water). The first four bulk compartments are considered as a combination of

subcompargnents of varying proportions of pure and particle phases. Equilibrium is assumed

to apply within each bulk compartment, but is not assumed between compartments.

Expressions for emissions, advective flows, degrading reactions, and inærphase transports by

diffusive and non-diffusive processes are included in CHEMFRANCE.The aim of this paper is basically to prcsent the theorctical foundations of CHEMFRANCE

and underline the difficulties encountered to construct such a model. For illustrative

purposes, the modeling of the environmental fate of isobutylene is presented.

MODEL DESCRIPTION

Regions of CHEMFRANCE

To develop CHEMFRANCE, it was necessary to divide France into a reasonable number of

regions of similar hydrogeological, climatic, and ecologic characteristics. These regions were

defined afær a careful study of the factors susceptible to have an effect on the environmental

fate of chemicals in France. Thus, it is obvious that all the economical or political parameters

inducing artifactitious divisions (e.g., limits of deparrnent, high density population area)

were excluded. The factors selected in priority were the drainage basins constituting the

French hydrographical network, the climate (i.e., precipitations, temPeratures, wind), the

nahrre of the soil, and the different qpes of vegetation.Under these conditions, France was divided into the twelve following regions (Figure l):

52

459

Region l- Bretagne, Pays de loireRegion 2- Nord, NormandieRegion 3- Bassin de ParisRegion 4- Alsace, LorraineRegion 5- Vallée du RhôneRegion 6- Massif centralRegion 7- AlpesRegion 8- Bassin méditerranéenRegion 9- Aquitaine, Midi-PyrénéesRegion 10- PyrénéesRegion 11- Bordure atlantiqueRegion 12- CentrcRegion 13 being FranceDividing a count

A Sandra - univ-lorraine.frdocnum.univ-lorraine.fr/public/UPV-M/Theses/1996/Bintein... · 2016. 9....

Documents

Transcript of A Sandra - univ-lorraine.frdocnum.univ-lorraine.fr/public/UPV-M/Theses/1996/Bintein... · 2016. 9....